Abnormal returns with momentum/

contrarian strategies using

exchange-traded funds
Received (in revised form): 25th February, 2008

Jack C. De Jong Jr.n

is an assistant professor at the School of Travel Industry Management at the University of Hawaii, as well as a CFA charterholder.
He received his PhD degree in International Management with an Asian Finance Specialisation from the Shidler College of
Business at the University of Hawaii.

S. Ghon Rhee
is the dean of the Sung Kyun Kwan University Business School in Seoul, Korea. He is on leave from the University of Hawaii Shidler
College of Business where he is the K.J. Luke Endowed Chair and Professor of International Finance and Banking. He received his
PhD degree in Finance from the Ohio State University. He is the executive director of the Asia-Pacific Financial Markets Research
Center at the University of Hawaii, as well as the managing editor of the Pacific-Basin Finance Journal.
n
School of Travel Industry Management, University of Hawaii, George Hall 112C, 2560 Campus Road, Honolulu, HI 96822, USA.
Tel: 1 808 956 5381; Fax: 1 808 956 5378; E-mail: jackj@hawaii.edu

Abstract Investing in portfolios of exchange-traded funds (ETFs) provides abnormal

returns that exceed transactions costs, when returns are adjusted for risk using Fama and
Frenchs three-factor model. Short formation and holding periods of one day to one week
provide abnormal contrarian returns, as past losers become winners and past winners
become losers. Medium formation and holding periods of 439 weeks provide abnormal
momentum returns, as past winners keep winning and past losers keep losing. Abnormal
returns for portfolios of ETFs result in an asset allocation setting including all four types of
ETFs, namely domestic, international, sector, and bond ETFs, with contrarian returns
maximised over a one-day formation and holding period, and with momentum returns
maximised over a 20-week formation and holding period.
Journal of Asset Management (2008) 9, 289299. doi:10.1057/jam.2008.27
Keywords: momentum abnormal returns, contrarian abnormal returns, exchange-traded
funds, ETF, market efficiency, anomaly

Introduction
This study extends the domain of
momentum/contrarianism to a relatively
new and popular investment vehicle, namely
exchange-traded funds or ETFs. ETFs are
powerful and flexible investment vehicles
that combine the diversified portfolio
features of mutual funds with the trading
possibilities of individual securities.
Currently, US ETFs function similarly to
passively managed index mutual funds, as
they are composed of a portfolio of stocks or
bonds that track a particular index, thus

providing diversification within the portion

of the market tracked by that index. The
four general categories of ETFs are (1)
broad-based domestic indices like the S&P
500, the NASDAQ 100, the Dow Jones US
Total Market, the Russell 3000, the Wilshire
5000, some style-specific indices in both a
value and a growth version, and size-based
indices including large cap, mid cap, small
cap, and micro cap; (2) sector indices
including consumer, energy, financial,
health, natural resources, real estate, utilities,
and technology; (3) international indices

& 2008 Palgrave Macmillan, 1470-8272 Vol. 9, 4, 289299 Journal of Asset Management
www.palgrave-journals.com/jam

289

De Jong and Rhee

including global stock indices, regional

indices, and country-specific indices; and
(4) bond indices including three of the
Lehman Treasury bond indices, two different
corporate bond indices, and the Lehman
TIPS index. What differentiates an ETF
from a mutual fund is that an ETF trades
on an exchange like a stock, enabling an ETF
to be purchased or sold at intraday market
prices, purchased on margin, sold short, and
traded via stop orders and limit orders.
Ordinary mutual funds can only be
purchased and sold by market orders for
end-of-day prices, and cannot be purchased
on margin or sold short, which prevents the
usual zero-investment momentum and
contrarian portfolios of buying the winners
and shorting the losers or of buying the losers
and shorting the winners. Also, many mutual
funds have redemption fees and other
constraints to discourage or prevent the
short-term trading necessary to implement a
momentum or contrarian strategy.
Figure 1 shows the growth in ETFs from
their introduction in 1993 with the SPDR
Trust Series tracking the S&P 500 Index to
2005s assortment of 217 ETFs consisting
of 80 broad-based domestic indices, 82
domestic sectors, 49 global/international
equities, and six bond ETFs. Clearly, ETFs
are on a growth path that should soon surpass
the dollar amount invested in equity index

mutual funds. From the Investment

Company Institutes (a mutual fund trade
organisation) December 2005 statistics, ETFs
(excluding Merrill Lynchs HOLDRS)
represent a market value of $300bn, which
represents over 5 per cent of $5.5tn invested
in stock and hybrid mutual funds.
Considering that about 10 per cent of stock
mutual fund investments are in indexed
investments as opposed to actively managed
funds, ETFs represent a significant portion
(almost 35 per cent) of the US wealth
invested in passively managed, index-type
investment vehicles. With the growing
popularity of ETFs among traders and
investors, such an innovative financial
product merits further study, especially
when it can generate abnormal returns via
a momentum or contrarian strategy.
Our research questions are (1) whether a
momentum investing strategy of buying
winners and shorting losers generates
abnormal returns in the ETF market,
(2) whether a contrarian investing strategy of
buying losers and shorting winners generates
abnormal returns in the ETF market, and
(3) which formation and holding period is
optimal for momentum investing and for
contrarian investing. Our results contribute
an affirmative answer to the first research
question about momentum, as buying the
winner decile of ETFs and shorting the loser

Number of ETFs

90
80

Domestic

70

Sector

60

Foreign
Bond

50
40
30
20
10
0
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
Years

Figure 1 Growth in ETFs

290

Journal of Asset Management Vol. 9, 4, 289299 & 2008 Palgrave Macmillan, 1470-8272

Abnormal returns with momentum/contrarian strategies

decile of ETFs provide statistically significant

abnormal returns for formation and holding
periods of 4, 8, 12, 16, 20, 26, and 39 weeks
with risk adjustment by Fama and Frenchs
(1993, 1996) three-factor model. The
annualised momentum abnormal returns
range from 8.4 to 13.5 per cent. Our results
contribute an affirmative answer to the
second research question about
contrarianism, as buying the loser decile of
ETFs and shorting the winner decile of ETFs
provide statistically significant contrarian
abnormal returns for formation and holding
periods of one day and one week. The
annualised contrarian abnormal returns range
from 19.8 to 86.9 per cent with risk
adjustment by Fama and Frenchs threefactor model. Following the classic approach
of Jegadeesh and Titman (1993), we find for
question (3) that a 20-week formation and
holding period provides the highest
annualised abnormal returns of 13.5 per cent
to an ETF momentum strategy, and a oneday formation and holding period provides
the highest annualised abnormal returns of
86.9 per cent to an ETF contrarian strategy.
These research questions are important,
because momentum and contrarianism are
both widespread anomalies identified by
researchers as well as investment strategies
used by investors to attempt to earn
abnormal returns.

Data and methodology

Following the Jegadeesh and Titman (1993)
methodology, we define the winner ETFs
as the top performing decile over various
formation periods and the loser ETFs as the
poorest performing decile over various
formation periods, and then form the
momentum portfolio that buys the winner
ETFs and shorts the loser ETFs over various
holding periods. Also, adapting the Chan
et al. (2000) methodology, we form the
winner-minus-loser (WML) portfolio each
week to increase the power of our tests; we
equally weight the appropriate winner and

loser ETFs in the portfolios formed each

week during the sample period and held for
the indicated amount of time. Also, with
ETFs tracking four different types of indices,
broad-based domestic, sector or industry,
foreign or country, and bond, we measure
momentum with all the various ETF types
pooled together to consider a strategy where
market performance may favour one of the
four types over the other three, which can
determine whether a momentum-based asset
allocation strategy exists, as well as to
maximise our sample size. Our sample
period runs from 20th March, 1996 to 31st
December, 2005, a period of 483 weeks,
with 19 ETFs available in 1996, so that the
top and bottom deciles begin with two ETFs
each as winners and losers, respectively. In
2005, 217 ETFs are available so that the
deciles of winners and losers both include 21
ETFs. We consider formation periods of 1
day, 1, 2, 4, 8, 12, 16, 20, 26, 39, and 52
weeks, coupled with holding periods of the
same length as the formation period for a
total of 11 different momentum strategies.
Clearly, the zero-investment contrarian
portfolio of buying the loser ETFs and
shorting the winner ETFs simply reverses the
sign of the zero-investment momentum
portfolio.
Since most ETFs are listed on the AMEX
and others are listed on the NYSE and
NASDAQ, the daily return data are available
on the Center for Research in Securities
Prices (CRSP) database. Most of the
previous studies use CRSP data, which is
of excellent quality when defining returns
using close to close security prices. To
evaluate the various ETFs in terms of risk
levels, we use Fama and Frenchs three-factor
model: RitRFt ai biRMRFt siSMBt
hiHMLt eit, where Rit is the return on
portfolio i in period t, RFt is the Treasury
bill rate in period t, RMRF is the excess
return on a value-weighted market proxy,
and SMB and HML are the returns on
zero investment, value-weighted, factormimicking portfolios for firm size and

& 2008 Palgrave Macmillan, 1470-8272 Vol. 9, 4, 289299 Journal of Asset Management

291

De Jong and Rhee

book-to-market, respectively. Daily data on

the factor-mimicking portfolios for the three
zero-investment factor-mimicking
portfolios, that is, RMRF, SMB, and HML,
are available on Kenneth R. Frenchs website
at: http://mba.tuck.dartmouth.edu/pages/
faculty/ken.french/data_library.html. Thus,
we calculate the cross-sectional as for the
excess return on the winner ETFs, the excess
return on the loser ETFs, as well as the
winner-minus-loser momentum portfolio to
measure abnormal positive or negative
returns. Although the three risk factors,
RMRF, SMB, and HML are all calculated
relative to US equities, no adjustments need
to be made to use the above factor risk
premiums for the foreign ETFs because all
the ETFs are traded in dollars, traded during
US market hours, and function as perfect
substitutes for the other three categories of
ETFs. Also, Zhong and Yang (2005) find that
the prices of international ETFs are greatly
influenced by US risk factors.

Main results
Table 1 shows that the annualised
momentum abnormal returns for WML
portfolios are statistically significant at
the 1 per cent level of significance, ranging
from 8.4 to 13.5 per cent, for strategies of
formation and holding periods from 4 to 39
weeks with risk adjusted by Fama and
Frenchs three-factor model. The annualised
momentum abnormal returns for the WML
portfolio are maximised at 13.5 per cent
for the 20-week formation and holding
period strategy. From the results for the
excess returns above the appropriate periodic
Treasury bill rate for the momentum winner
and loser ETF portfolios, we find that the
losers drive the WML risk-adjusted results,
with loser annualised momentum abnormal
returns being very significant at the 1 per
cent level for all formation and holding
periods from 4 to 39 weeks with magnitudes
ranging from 8.9 to 12.8 per cent.
Annualised contrarian abnormal returns
are statistically significant at the 1 per cent

292

level of significance at 86.9 per cent for the

short formation and holding periods of one
day and at the 5 per cent level of significance
at 19.8 per cent for the formation and holding
periods of one week with risk adjusted by
Fama and Frenchs three-factor model. Thus,
by reversing our momentum strategy to a
contrarian strategy of buying the past losers
and shorting the past winners, statistically
significant returns are possible with abnormal
annualised loser returns for the one-day
formation and holding period of 45.4 per
cent, and abnormal annualised winner returns
for the one-day formation and holding period
of 41.6 per cent. Thus, a formation and
holding period of one day maximises our
contrarian abnormal returns, while a
formation and holding period of 20 weeks
maximises our momentum abnormal return.1

Abnormal returns net of

transaction costs
Supporters of rational, efficient markets
question whether the momentum and
contrarian abnormal returns are real or
illusory after considering transactions costs,
since both momentum and contrarian
strategies require a significant amount of
trading to implement, potentially costing the
investor or arbitrageur the bidask spread,
brokerage commissions, and price impact for
large orders. Lesmond et al. (2004)
characterise the momentum profits with
individual stocks identified by Jegadeesh and
Titman (1993, 2001) and Hong et al. (2000)
as illusory because the momentum profits net
of their transactions cost estimates are
insignificantly different from 0. ETFs,
however, are much less costly to trade than
individual equities with smaller bidask
spreads and more liquidity to reduce the
price impact of large trades.
Table 2 shows the momentum returns net
of transactions costs for the formation and
holding period of 26 weeks, which is the
usual recommended momentum strategy in
most previous studies. On average, the

Journal of Asset Management Vol. 9, 4, 289299 & 2008 Palgrave Macmillan, 1470-8272

1day 1day

1wk 1wk

2wk 2wk

4wk 4wk

8wk 8wk

12wk 12wk

16wk 16wk

20wk 20wk

26wk 26wk

39wk 39wk

52wk 52wk

Fama and Frenchs three-factor model: RitRFt=ai+biRMRFt+siSMBt+hiHMLt+eit

WML
Constant (%) 86.9250
19.7964
11.0838
11.6922
13.0897
p-value
0.0000
0.0213
0.1056
0.0164
0.0006

10.2860
0.0019

13.2135
0.0000

13.5426
0.0000

12.8348
0.0000

8.4255
0.0083

0.0062
0.9985

Winner
Constant (%)
p-value

41.5500
0.0000

14.6744
0.0264

2.4336
0.5395

0.1768
0.9543

0.3237
0.8435

0.3315
0.8364

3.5321
0.0292

4.6792
0.0015

3.8318
0.0019

0.6001
0.5465

3.9998
0.0344

Loser
Constant (%)
p-value

45.3750
0.0000

7.5140
0.2342

8.6502
0.0981

11.5154
0.0020

12.7660
0.0001

10.6175
0.0002

9.6811
0.0001

8.8634
0.0002

9.0030
0.0002

9.0256
0.0008

3.9937
0.1332

Notes: (1) Annualised abnormal returns for the period indicated. Periodic abnormal returns, that is, alpha intercepts, are annualised by multiplying by either 250 trading days for
one day one day strategy or by 52 divided by the number of weeks in the 10 weekly strategies.
(2) Formation period includes daily returns from Thursday to Wednesday; holding period includes daily returns from Friday to Thursday.
(3) The winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation
period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the
indicated period of time.
(4) The WML is the zero net- investment portfolio created by buying the winner ETFs and by shorting the loser ETFs for the indicated holding period.
(5) p-Values are computed with robust standard errors corrected for heteroskedasticity and autocorrelation using the NeweyWest adjustment (1987).

Abnormal returns with momentum/contrarian strategies

& 2008 Palgrave Macmillan, 1470-8272 Vol. 9, 4, 289299 Journal of Asset Management

Table 1 Annualised abnormal returns: Exchange-traded funds and momentum returns: 21st March, 199631st December, 2005

293

De Jong and Rhee

winner ETFs consisted of 10.64 per cent

domestic and bond ETFs, 46.52 per cent
sector ETFs, and 42.84 per cent international
ETFs, while the loser ETFs consisted of
8.08 per cent domestic and bond ETFs,
54.97 per cent sector ETFs, and 36.95 per
cent international ETFs. Quoted bidask
spreads and brokerage commissions total 8.25
per cent per year, which reduced the net
annualised abnormal returns to 4.58 per cent
with risk adjustment by Fama and Frenchs
three-factor model. Huang and Wei (2004)
estimated the effective bidask spread to be
about 30 per cent less than the quoted bid
ask spread, as transactions often take place
between the quoted bid and ask prices; using
effective bidask spreads reduces the
transactions cost to 5.82 per cent per year.
With effective spreads, the net annualised
abnormal returns are 7.01 per cent with
risk adjustment by Fama and Frenchs
three-factor model. Clearly, the momentum
abnormal returns earned using ETFs are not
illusory, but represent economically viable
and statistically significant returns even when
reduced by transactions costs.

Results due to portfolio

rebalancing
Another issue is portfolio rebalancing. The
usual Jegadeesh and Titman methodology
rebalances the winner and loser portfolios
monthly, resulting in the six-month winners
and losers, consisting of six equally weighted
portfolios with one-sixth formed from
todays winners and losers and one-sixth
formed from each of the previous one, two,
three, four, and five months winners and
losers. Our methodology uses equal
weightings of the top and bottom decile ETF
performers to form the winner and loser
portfolios like Jegadeesh and Titman, but
forms a new portfolio each week during the
sample period and holds it for the indicated
period of time without rebalancing.
Table 3 compares the annualised
abnormal returns for portfolios with a
formation and holding period of 26 weeks,
using our methodology with no rebalancing
as well as with one-week rebalancing, and
portfolios with a formation and holding
period of 24 weeks with four-week
rebalancing similar to Jegadeesh and Titman.

Table 2 Annualised abnormal returns net of transactions costs for WML with 26-week formation and
holding period: Exchange-traded funds and momentum returns: 21st March, 199631st December, 2005
Model

Annualised
returns (%)

Transactions
costs (%)

Net annualised
returns (%)

Fama and Frenchs three-factor model

12.83

8.25

4.58

Notes: (1) Annualised abnormal returns for the period indicated. Periodic abnormal returns, that is, alpha
intercepts, are annualised by multiplying by 2 to convert the 26-week abnormal return to an annualised abnormal
return.
(2) Formation period includes daily returns from Thursday to Wednesday; holding period includes daily returns from
Friday to Thursday.
(3) The winners represent the top decile of ETF returns available during the formation period; the losers represent
the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with
equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held
for the indicated period of time.
(4) The WML is the zero net-investment portfolio created by buying the winner ETFs and by shorting the loser ETFs
for the indicated holding period.
(5) Actual transactions were tabulated by domestic, sector, or international ETFs over the 9.29 years studied.
(6) Quoted bid ask spreads were estimated as the higher of those identified in Huang and Wei (2004)and Salomon
Smith Barney (2002) resulting in estimates of 0.33 per cent for domestic ETFs, 0.62 per cent for sector ETFs, and
0.867 per cent for international ETFs.
(7) Commissions were estimated at 0.13 per cent by combining Scottrades $7.00 per trade with an estimated
account balance of $125,000 invested long in the winner ETFs and $125,000 invested short in the loser ETFs.
Scottrades flat commission applies to both market and limit orders regardless of trade frequency, account
balance, or number of shares in the transaction.

294

Journal of Asset Management Vol. 9, 4, 289299 & 2008 Palgrave Macmillan, 1470-8272

Abnormal returns with momentum/contrarian strategies

Table 3 Annualised abnormal returns using Fama and Frenchs three-factor model exchange-traded funds
and momentum returns: 21st March, 199631st December, 2005
26wk 26wk
No rebalancing

26wk 26wk
1wk rebalancing

Fama and Frenchs three-factor model: RitRFt=ai+biRMRFt+siSMBt+hiHMLt+eit

WML
Constant (%)
12.8348
16.0264
p-value
0.0000
0.0728

24wk 24wk
4wk rebalancing

16.1109
0.0935

Winner
Constant (%)
p-value

3.8318
0.0019

2.7248
0.5230

1.7641
0.7150

Loser
Constant (%)
p-value

9.0030
0.0002

13.3016
0.1033

14.3468
0.0689

Notes: (1) Annualised abnormal returns for the period indicated. Periodic abnormal returns, that is, alpha
intercepts, are annualised by multiplying by 2 for the no rebalancing or by 52 divided by the number of weeks in
the rebalancing period.
(2) Formation period includes daily returns from Thursday to Wednesday; holding period includes daily returns from
Friday to Thursday.
(3) The winners represent the top decile of ETF returns available during the formation period; the losers represent
the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with
equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held
for the indicated period of time for the no rebalancing.
(4) The WML is the zero net-investment portfolio created by buying the winner ETFs and by shorting the loser ETFs
for the indicated holding period.
(5) The 26 weeks with one-week rebalancing portfolios represent 26 equally weighted portfolios formed from
the winner and loser ETFs from today, one week ago, two weeks ago, three weeks ago, and each other week ago
up to and including 25 weeks ago. Thus, the winner and loser returns represent one-week returns.
(6) The 24 weeks with four-week rebalancing portfolios represent six equally weighted portfolios formed from
the winner and loser ETFs from today, four weeks ago, eight weeks ago, 12 weeks ago, 16 weeks ago, and
20 weeks ago. Thus, the winner and loser returns represent four-week returns.
(7) p-values are computed with robust standard errors corrected for heteroskedasticity and autocorrelation using
the NeweyWest adjustment (1987).

Rebalancing six portfolios over four-week

periods necessitates using a 24-week
formation and holding period rather than a
26-week formation and holding period. The
results are qualitatively the same for all three
methodologies of rebalancing, with similar
annualised abnormal return magnitudes
of 12.8 per cent with no rebalancing vs
16.0 per cent with one-week rebalancing
vs 16.1 per cent with four-week rebalancing,
while no rebalancing has greater statistical
significance than both one- and four-week
rebalancing.

Results due to the type of ETF

Our results so far have not identified the
source of the abnormal momentum or

contrarian returns because we formed all

winner and loser portfolios of ETFs from the
asset pool that included all four types of
ETFs, namely domestic, sector,
international, and bond. Our study
purposefully chooses all four types of ETFs
simultaneously to extend momentum/
contrarian studies to the asset allocation
domain as well as to maximise the length of
the sample period studied, since ETFs are a
relatively new investment vehicle. On
average, about 51 per cent of the ETFs in the
winner and loser portfolios are sector ETFs
and about 40 per cent are international ETFs,
with the remaining 9 per cent from the
domestic and bond ETFs. To control for the
effects of including sector and international
ETFs, we augmented Fama and Frenchs

& 2008 Palgrave Macmillan, 1470-8272 Vol. 9, 4, 289299 Journal of Asset Management

295

De Jong and Rhee

three-factor model with dummy variables,

identifying whether or not the winner, loser,
and WML portfolios included sector or
international portfolios.
Table 4, in the Including dummies
columns, clearly shows large and very
significant abnormal momentum returns for
WML portfolios formed from domestic and
bond ETFs only with annualised abnormal
momentum returns ranging from 44.7 per
cent with no rebalancing to 70.9 per cent
with four-week rebalancing. These large
abnormal momentum returns are driven by
the loser portfolios of domestic and bond
ETFs, which continue to lose a very
significant annualised abnormal momentum
return of 47.6 per cent with no rebalancing
to 56.6 per cent with four-week
rebalancing. Clearly, the economic times
when the top and bottom 26-week

performance deciles are dominated by

domestic and bond ETFs rather than the
more volatile and focused sector and
international ETFs are excellent times to
implement a momentum strategy for the
next 26 weeks. At more typical economic
times when the top and bottom 26-week
performance deciles include either sector or
international ETFs or both, the abnormal
momentum returns over the next 26 weeks
are reduced, with the result being driven by
the sector and international ETF losers losing
significantly less than their domestic and
bond ETF counterparts.
Table 5 frames the analysis of asset
allocation vs type of ETF differently, by
comparing the 26-week formation and
holding period WML, winner, and loser
ETF performance over a shorter sample
period, from 1st January, 1999 to 31st

Table 4 Abnormal returns using Fama and Frenchs three-factor model exchange-traded funds and
momentum returns: 21st March, 199631st December, 2005
26wk 26wk
No rebalancing

WML
Constant
p-value
Mkt RF
p-value
HML
p-value
SMB
p-value
DumSect
p-value
DumInt
p-value
Adjusted R2
Winner
Constant
p-value
Mkt RF
p-value
HML
p-value
SMB
p-value
DumSect
p-value
DumInt
p-value
Adjusted R2

296

24wk 24wk
4wk rebalancing

No dummies

Including dummies

No dummies

Including dummies

0.0642
0.0000
0.2173
0.1044
0.6994
0.0000
0.5045
0.0058

0.2234
0.0000
0.5151
0.0000
0.9840
0.0000
1.1623
0.0000
0.2585
0.0000
0.0022
0.9587
0.2930

0.0124
0.0935
0.3915
0.0273
0.6155
0.0082
0.5375
0.0450

0.0545
0.0037
0.4279
0.0168
0.6431
0.0097
0.6535
0.0193
0.0321
0.1039
0.0217
0.2752
0.1544

0.0112
0.4922
0.9595
0.0000
0.1938
0.0000
0.4317
0.0000
0.0042
0.8057
0.0341
0.0415
0.5705

0.0014
0.7150
1.1392
0.0000
0.2606
0.0368
0.3357
0.0027

0.1431
0.0192
0.0019
0.9686
0.0000
0.1675
0.0003
0.4006
0.0000

0.5696

0.1343

0.6301

0.0145
0.2714
1.1450
0.0000
0.3157
0.0317
0.3739
0.0049
0.0026
0.7996
0.0184
0.1556
0.6294

Journal of Asset Management Vol. 9, 4, 289299 & 2008 Palgrave Macmillan, 1470-8272

Abnormal returns with momentum/contrarian strategies

Table 4 Continued
26wk 26wk
No rebalancing

Loser
Constant
p-value
Mkt RF
p-value
HML
p-value
SMB
p-value
DumSect
p-value
DumInt
p-value
Adjusted R2

24wk 24wk
4wk rebalancing

No dummies

Including dummies

No dummies

Including dummies

0.0450
0.0002
1.1859
0.0000
0.5319
0.0000
0.1039
0.4983

0.2380
0.0000
1.3995
0.0000
0.8358
0.0000
0.7177
0.0000
0.2444
0.0000
0.0569
0.0033
0.5391

0.0110
0.0689
1.5307
0.0000
0.3549
0.0407
0.2018
0.3483

0.0435
0.0037
1.5561
0.0000
0.4209
0.0118
0.3241
0.1181
0.0329
0.0262
0.0135
0.1019
0.6115

0.4279

0.5954

Notes: (1) Augmented Fama and Frenchs three-factor model with dummy variables for sector and international
ETFs:
Rit  RFt ai bi RMRFt si SMBt hi HMLt
ci DumSt di DumIt eit
where DumSt=1 if the portfolio at time t includes at least one sector ETF, and 0 otherwise; and DumIt=1 if the
portfolio at time t includes at least one international ETF, and 0 otherwise.
(2) Dependent variable is periodic return for the WML which is the winner-minus-loser portfolio, a zero
net-investment portfolio formed by buying the winners and shorting the losers over the listed formation and holding
periods, or excess periodic return for the winner or loser portfolio above the risk free rate as proxied by the
appropriate periodic Treasury bill rate over the listed formation and holding period.
(3) Formation period includes daily returns from Thursday to Wednesday; holding period includes daily returns from
Friday to Thursday.
(4) The winners represent the top decile of ETF returns available during the formation period; the losers represent
the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with
equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held
for the indicated period of time for the no rebalancing.
(5) The 24 weeks with four-week rebalancing portfolios represent six equally weighted portfolios formed from the
winner and loser ETFs from today, four weeks ago, eight weeks ago, 12 weeks ago, 16 weeks ago, and 20 weeks
ago. Thus, the winner and loser returns represent four-week returns.
(6) RMRF is the excess return on the market portfolio, which is the value-weighted return on all NYSE, AMEX, and
NASDAQ stocks from CRSP minus the periodic Treasury bill rate (from Ibbotson Associates) over the listed
formation and holding periods.
(7) HML (high book-to-market minus low book-to-market) is the average return on the two value portfolios minus
the average return on the two growth portfolios, that is, HML=1/2 (small value+big value)1/2 (small growth+big
growth).
(8) SMB (small size minus big size) is the average return on the three small portfolios minus the average return
on the three big portfolios, SMB=1/3 (small value+small neutral+small growth)1/3 (big value+big neutral+big
growth).
(9) DumSect is a dummy variable that equals 1 if the WML, winner, or loser portfolio includes at least one sector
ETF, and it equals 0, otherwise.
(10) DumInt is a dummy variable that equals 1 if the WML, winner, or loser portfolio includes at least one
international ETF, and it equals 0, otherwise.
(11) p-values are computed with robust standard errors corrected for heteroskedasticity and autocorrelation using
the NeweyWest adjustment (1987).

December, 2005, to allow a sufficient

number of ETFs in each category to
meaningfully define winners and losers.
With risk adjustment by Fama and Frenchs
three-factor model, the asset allocation pool
including all four types of ETFs generates a

somewhat significant abnormal contrarian

return, but the bond ETFs generate a
significant abnormal momentum return and
the domestic, international, and sector ETFs
generate abnormal contrarian returns that are
insignificantly different from 0. Among the

& 2008 Palgrave Macmillan, 1470-8272 Vol. 9, 4, 289299 Journal of Asset Management

297

De Jong and Rhee

Table 5 Abnormal returns using Fama and Frenchs three-factor model with 26-week formation and
holding periods exchange-traded funds and momentum returns: 1st January, 199931st December, 2005
Types of ETFs included
All types

Bond

Domestic

International

Sector

WML
Constant
p-value
Mkt RF
p-value
HML
p-value
SMB
p-value
Adjusted R2

0.0169
0.0880
0.5199
0.0000
0.5668
0.0000
1.2073
0.0000
0.4194

0.0177
0.0176
0.0160
0.8779
0.0597
0.6054
0.2130
0.1134
0.0872

0.0077
0.3985
0.4614
0.0000
0.4910
0.0000
1.3176
0.0000
0.3951

0.0065
0.4869
0.0980
0.1951
0.0781
0.2383
0.2713
0.0296
0.0351

0.0230
0.1296
0.6963
0.0000
0.9995
0.0000
1.6787
0.0000
0.4223

Winner
Constant
p-value
Mkt RF
p-value
HML
p-value
SMB
p-value
Adjusted R2

0.0071
0.1979
1.1181
0.0000
0.2284
0.0000
0.4002
0.0001
0.7923

0.0098
0.1467
0.3147
0.0012
0.4040
0.0000
0.4391
0.0012
0.2491

0.0034
0.6395
0.9048
0.0000
0.2149
0.0069
0.6932
0.0001
0.6548

0.0195
0.0188
1.2354
0.0000
0.2336
0.0001
0.4502
0.0013
0.6629

0.0016
0.8314
1.2738
0.0000
0.3492
0.0000
0.1982
0.1333
0.6570

Loser
Constant
p-value
Mkt RF
p-value
HML
p-value
SMB
p-value
Adjusted R2

0.0240
0.0018
1.6380
0.0000
0.3384
0.0000
0.8070
0.0000
0.8229

0.0276
0.0000
0.3307
0.0000
0.4636
0.0000
0.2261
0.0002
0.2818

0.0110
0.0060
1.3662
0.0000
0.2761
0.0001
0.6244
0.0000
0.8384

0.0260
0.0001
1.1374
0.0000
0.1556
0.0104
0.1789
0.1060
0.6648

0.0247
0.0433
1.9700
0.0000
0.6503
0.0000
1.4805
0.0000
0.7460

Notes: (1) Fama and Frenchs three-factor model: RitRFt=ai+biRMRFt+siSMBt+hiHMLt+eit.

(2) Dependent variable is periodic return for the WML which is the winner-minus-loser portfolio, a zero
net-investment portfolio formed by buying the winners and shorting the losers over the listed formation and holding
periods, or excess periodic return for the winner or loser portfolio above the risk free rate as proxied by the
appropriate periodic Treasury bill rate over the listed formation and holding period.
(3) Formation period includes daily returns from Thursday to Wednesday; holding period includes daily returns from
Friday to Thursday.
(4) The winners represent the top decile of ETF returns available during the formation period; the losers represent
the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with
equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held
for the indicated period of time.
(5) All sample periods include 1st January, 199931st December, 2005 except for the bond ETFs which include the
sample period from 27th July, 2002 to 31st December, 2005.
(6) RMRF is the excess return on the market portfolio, which is the value-weighted return on all NYSE, AMEX, and
NASDAQ stocks from CRSP minus the periodic Treasury bill rate (from Ibbotson Associates) over the listed
formation and holding periods.
(7) HML (high book-to-market minus low book-to-market) is the average return on the two value portfolios minus
the average return on the two growth portfolios, that is, HML=1/2 (small value+big value)1/2 (small growth+big
growth).
(8) SMB (small size minus big size) is the average return on the three small portfolios minus the average return on
the three big portfolios, SMB=1/3 (small value+small neutral+small growth)1/3 (big value+big neutral+big
growth).
(9) p-values are computed with robust standard errors corrected for heteroskedasticity and autocorrelation using
the NeweyWest adjustment (1987).

298

Journal of Asset Management Vol. 9, 4, 289299 & 2008 Palgrave Macmillan, 1470-8272

Abnormal returns with momentum/contrarian strategies

winners, only the international ETFs

generate a positive significant abnormal
return. The bond loser ETFs generate a
negative, very significant abnormal return,
but the loser ETFs for the asset allocation
pool and the domestic, international, and
sector generate positive, very significant
abnormal returns. Since losing 19961998
from the sample reduces the magnitude and
significance of the abnormal momentum/
contrarian returns across all five ETF pools,
some of the advantage of the asset allocation
pool of all four types of ETFs reported earlier
is due to the longer available sample period.2
This analysis, however, indicates some
differences between the holding period
performances of the four different types of
ETFs, which may lead to improvements over
a nave momentum/contrarian strategy.

Conclusion
This study extends Jegadeesh and Titmans
momentum/contrarian anomaly to a new
domain: portfolios of ETFs that either buy
the winners and short the losers or buy the
losers and short the winners, respectively.
Currently, all US ETFs are passively managed
to track an index, not actively managed to
time the market or beat the market by
loading up on high momentum stocks. Yet,
in spite of this disadvantage to actively
managed mutual funds, ETFs provided
economically and statistically significant
abnormal returns to contrarian strategies of
buying the loser ETFs and shorting the
winner ETFs with formation and holding
periods of one day and one week, and to
momentum strategies of buying the winner
ETFs and shorting the loser ETFs with
formation and holding periods from 4 to 39
weeks. This study is also the first to
demonstrate momentum in a changing asset
allocation setting that includes US stocks, US
bonds, foreign stocks, as well as sector or
industry funds. In contrast to Lesmond et al.
(2004), we find that momentum/contrarian
abnormal returns are not illusory, but are

achievable by investors and arbitrageurs using

ETFs, which are ideal instruments with
which to implement such a contrarian or
momentum strategy.
Notes
1. Results with no risk adjustment are qualitatively the same
as those with risk adjustment by Fama and Frenchs threefactor model. Results with extra time between the
formation and holding period are also qualitatively the
same, with a slight reduction in the magnitude of the
abnormal returns for most strategies.
2. With no risk adjustment, the asset allocation pool
including all four types of ETFs generates very
significant abnormal momentum returns for the WML
portfolio, as do both the domestic only and sector only
WML portfolios. The sector ETFs and domestic ETFs
generate larger abnormal momentum returns for the
WML portfolio than the asset allocation pool, but the
bond ETFs and international ETFs both generate
smaller, less statistically significant abnormal returns for the
WML portfolio than the asset allocation pool, primarily
because both the bond loser abnormal returns and the
international loser abnormal returns are positive rather
than negative.

References
Chan, K., Hameed, A. and Tong, W. (2000) Profitability of
Momentum Strategies in the International Equity
Markets, Journal of Financial and Quantitative Analysis,
35(2), 153172.
Fama, E. F. and French, K. R. (1993) Common Risk Factors
in the Returns on Stocks and Bonds, Journal of Financial
Economics, 33(1), 356.
Fama, E. F. and French, K. R. (1996) Multifactor
Explanations of Asset Pricing Anomalies, Journal of
Finance, 51(1), 5584.
Hong, H., Lim, T. and Stein, J. C. (2000) Bad News Travels
Slowly: Size, Analyst Coverage, and the Profitability of
Momentum Strategies, Journal of Finance, 55(1), 265295.
Huang, C. H. and Wei, P. (2004) Bid Ask Spreads and
Holding Periods of Exchange Traded Funds, Working
Paper.
Jegadeesh, N. and Titman, S. (1993) Returns to Buying
Winners and Selling Losers: Implications for Stock Market
Efficiency, Journal of Finance, 48(1), 6591.
Jegadeesh, N. and Titman, S. (2001) Profitability of
Momentum Strategies: An Evaluation of Alternative
Explanations, Journal of Finance, 56(2), 699720.
Lesmond, D. A., Schill, M. J. and Zhou, C. (2004) The
Illusory Nature of Momentum Profits, Journal of Financial
Economics, 71(2), 349380.
Salomon Smith Barney (2002) Snapshot Study Performed 40
Random Snapshots of all ETFs in January 2002, Working
Paper.
Zhong, M. and Yang, H. (2005) Risk Exposures and
International Diversification: Evidence from iShares,
Journal of Business Finance & Accounting, 32(34), 737771.

& 2008 Palgrave Macmillan, 1470-8272 Vol. 9, 4, 289299 Journal of Asset Management

299

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.